Adjusted Stats 2009 Year in Review Part VII

This is the seventh in a series of posts that will review the 2009 season (as well as 2008) from an adjusted stats perspective.

Tentative Publication Schedule

Part I: Which Stats Correlate Best to Winning? - 6/24/10
Part II: Drilling Down and Regression Models - 6/28/10
Part III: Passing Efficiency Formula - 6/30/10
Part IV: Conference Strengths and Pace - 7/2/10
Part V: Testing Conventional Wisdom - 7/16/10
Part VI: Team Matchups - 7/20/10
Part VII: Year-to-Year Changes - 7/22/10
Part VIII: Points per Yard Efficiencies - 7/26/10
Part IX: Data Dump - Team Rankings - 7/28/10
Part X: Data Dump - Team Reports - 7/30/10

Year-to-Year Changes

The team reports that will be dumped on you next week include data from both 2008 and 2009 along with a column detailing the year-to-year change for each team in each statistic. With that information we can see how a team improved or got worse at any phase of the game for which we have data. Some statistics we know are due almost entirely to skill and execution, such as total rushing per carry. But some statistics seem to have a lot to do with luck, such as percentage of forced fumbles recovered. There are some seasons where a team seems to recover far more than their share of fumbles either offensively or defensively. The percentage of fumbles recovered by the defense should always end up near 50% (it was 50.2% nationally in 2009 and 49.1% in 2008), something I’m confident of to the point that I don’t run an adjusted stat for this.

But should I? Although I don’t have an adjusted stat available to run the study, the raw percentage of forced fumbles recovered stat should serve us just fine here. If a defense truly swarms to the ball more than the average team to the point that we should expect them to consistently recover more fumbles than other teams, that should show up in the raw stat. A coefficient of 0.116 exists for correlation between recovered fumble percentage in 2008 and the same in 2009. This is a positive correlation but falls below the 0.200 statistical significance threshold that we’ve been using. It is a higher value, though, than the 0.053 coefficient for adjusted forced fumbles per carry from 2008 to 2009. Clearly a team’s performance when it comes to fumbles can’t be expected to remain consistent from year-to-year. Here is a table showing both offensive and defensive fumble-related correlations from 2008 to 2009:

It appears that the coachspeak about emphasizing ball protection for a team’s offense is well-founded. The strongest correlation is for the frequency with which a team fumbles the ball. On the other hand, coachspeak about emphasizing stripping the ball and forcing fumbles appears to be largely without merit. The frequency with which a defense forces fumbles is statistically non-correlated from one year to the next. Fumble recovery percentage shows some possible correlation but not high enough to be certain; perhaps the correlation that does exist is linked to run/pass ratio for an offense. Fumbles after receptions are more frequently recovered by the defense, but fumbles after sacks are more frequently recovered by the offense (note: those are assumptions), so it may be related to pass protection in combination with playcalling ratio. Regardless, looking at the defensive statistics above it’s clear that you should not expect a team that performs either extremely well or extremely poorly in forcing and recovering fumbles in one season to repeat that performance the next year.

Before proceeding to cover all the year-to-year correlations for tracked statistics let’s take a look at the rest of the turnover-related statistics:

What to make of these results? In the fumble section above, I gave credit to coaches and teams that emphasize protecting the ball. But here we see that throwing interceptions shows basically zero correlation from one year to the next. It could be a case of quarterbacks being replaced if they throw too many interceptions and that interceptions are more directly related to individual execution than fumbles, which may be somewhat chaotic or random. What we do see, though, is that while defenses are not consistent in their year-to-year ability to cause or recover fumbles they do show more consistency when it comes to intercepting opposition pass attempts.

So there’s a lot to consider when we’re looking at year-to-year correlations. If a statistic can be easily influenced by a single player then we will likely see lower correlation coefficients. If there’s a significant randomness factor contributing to the statistic we will also see weaker correlations. And with the significant annual roster turnover in college sports how large can we truly expect any coefficient to be? The next page shows a table detailing the year-to-year coefficients for all other available adjusted stats. The first two tables showed a decent enough level of agreement between the raw and adjusted correlations (with the adjusted numbers appropriately showing greater correlation in most cases) that the adjusted will be used moving forward.

What should we expect? Based on the above logic we would expect stats obtained by the full team to show the most correlation as roster turnover should affect the entire team less drastically than it may affect individual positions. Taking the next logical step we would anticipate that full unit statistics, i.e., overall offensive and defensive statistics, to show stronger correlation than more specific stats. And as you’ll see while reviewing the table that logic holds fairly steady. But there’s something else made obvious when looking at the results that I hadn’t considered before running the numbers. Take a look at the table and come back.

The clear trend is that defensive performance correlates better from year-to-year than offensive performance. What are the possible explanations and associated strategic implications? Above I posited that individual focus can decrease year-to-year correlation. And the defensive side of the ball is less prone to being dominated by individual players than the offensive side. On offense, one player at any time possesses the ball and that player will have a greater impact on his unit’s performance than his teammates.

A second possible explanation is that offenses are more prone to change overall strategy and tactics on a yearly basis than defenses. The variety of defensive strategy is lower than offensive, which should cause greater stability on the defensive side of the ball and therefore stronger year-to-year correlations. Anecdotally it seems that the game of cat-and-mouse is usually led by offensive innovation from a few teams and then others will follow while the strategic defensive responses spread sooner throughout the country if they show success against the latest offensive attacks.

So now here's the really long table showing the year-to-year correlation results for the rest of the adjusted stats in my database:

Given that defensive performance is more consistent from season to season, it would seem that identifying and keeping a solid defensive coordinator and staff would be extremely important to a program. The program can be confident that the defensive performance is more likely to stay above par on an annual basis and that any difficulties on the offensive side of the ball can be addressed more quickly via personnel or strategic changes. So should teams look for defensive-minded head coaches versus offensive-oriented leaders? That could go either way because if a school can find an offensive head coach that can deliver consistent performance on that side of the ball then they will have a huge leg up on the competition. Thoughts?